We have shown that it is not currently possible to use the Cochrane Reviews of trials to examine the impact of smoking cessation interventions delivered in primary care on inequalities in health by socioeconomic status. None of the included Cochrane Reviews considered if the effectiveness of trials of smoking cessation interventions differed by SES. Within reviews we found that SES is rarely reported in randomised trials of smoking cessation interventions delivered in primary care. None of the three studies that reported SES in relation to outcome found an association. A number of different measures of SES are used. Where reported, measures are used to describe the sample at baseline or to demonstrate similarity of randomised groups rather than to consider if effectiveness differs by SES.
Comparison To Existing Literature
To the best of our knowledge, this is the first study to have considered the reporting of SES in trials of smoking cessation interventions delivered in primary care. Our findings are similar to those found in systematic reviews of other interventions delivered in primary care. The first was a systematic review of RCTs of primary care-based physical activity interventions where only three out of 171 RCTs reported a measure of SES (65). Two of these three trials found no difference in effectiveness by SES; the remaining study did not provide a clear indication of whether a differential effect was present. The second, a systematic review of primary care interventions to aid management of multimorbidity found that none of the ten studies analysed effects by SES, while three out of the ten reported SES at baseline (66).
Together, our findings, alongside previews reviews, suggest that there is insufficient trial evidence to consider if interventions delivered in primary care are equitable in design, implementation, accessibility, uptake and outcome. The low reporting of SES in trials could be due to trial reporting guidance documents, such as CONSORT 2010 statement (67), not considering equity. However, this limitation is counteracted somewhat by the extension of the CONSORT statement in 2017 with CONSORT-Equity (68). Since the latest study featured in the synthesis in this thesis was published in 2016, it is not possible to consider if the addition of CONSORT-Equity 2017 has improved the reporting of SES in trials, although recommendations about this have been around for some time (69).
The lack of SES data may be due to the exclusion of SES from pre-selected subgroup analyses. The studies identified in this review that did consider the variation of effectiveness of smoking cessation interventions by SES found no association, possibly due to the small number of participants in each group. This supports evidence from Inglis et al., (70) suggesting that subgroup analysis by SES often lacked methodological credibility. Given the time and resources involved with conducting clinical trials, it is unlikely that single studies are going to be sufficiently powered to avoid type II errors when considering the impact of interventions on inequalities.
Other systematic reviews have considered equity of smoking cessation interventions using data in addition to those from randomised controlled trials. Brown et al., considered the equity impact of European individual-level smoking cessation interventions (30). They found that NHS stop smoking services in the UK achieve equitable effects, and services which specifically targeted low-SES smokers achieved a higher uptake among low-SES smokers, which may attenuate the inequality gradient that exists for smoking cessation (14). This was supported in a more recent systematic review that focused solely on the UK which found that low-SES smokers were more likely to receive GP brief intervention or referral, but low-SES smokers were less successful at smoking abstinence (29). However, both reviews use a much broader definition of SES compared to this study and the other two previously discussed systematic reviews (65, 66) by including education level and occupation within the description of SES. The other reviews, and the PROGRESS-Plus criteria, separate education and occupation from SES.
Two trials highlighted in our study targeted individuals with low SES (50, 51), with evidence being mixed as to whether targeting interventions increases abstinence in this population. This targeting of interventions to those of a low-SES group was considered more extensively in a systematic review from Bull et al., (71). The review considered the effectiveness of targeted behavioural interventions for smoking cessation, physical activity and healthy eating. Post-intervention, there was a small but significant positive effect but this was not sustained in the longer term for smoking cessation or physical activity. This was supported by a rapid review from Vilhelmsson and Östergren (72) which found “no support for the notion that the methods used to reduce smoking decrease inequalities in health”, and limited evidence that targeted interventions decrease inequalities in diet or physical activity. Consequently, future research should consider if targeted interventions for smoking cessation are effective, and this research needs to be more strongly theoretically driven than is currently the case.
Strengths and Limitations
Our search strategy was effective within the time and resource limitations available. Using Cochrane reviews, which only include high-quality RCTs, meant that we could highlight relevant literature that had been identified using comprehensive searching and evidence appraisal. Therefore, it is unlikely that we missed relevant research by only focusing on trials included in Cochrane reviews. We had regular discussions with topic experts which informed our methodological approach and allowed for greater consideration of the implications and the context of our findings.
The main limitation that arises from this approach is that our study is limited to literature published prior to the conducting of the searches by the authors of the Cochrane reviews. Hence, trials published since the conducting of each individual Cochrane review (date range 2013–2019) that meet each review’s inclusion criteria post-hoc are not included. However, focusing only on RCTs meant that potentially relevant data from observational and other experimental studies were excluded. A further limitation of just using trials included in Cochrane reviews is the inconsistencies that exist in inclusion and exclusion criteria across the identified reviews. For example, differing restrictions were made in terms of study populations – some restricted trial eligibility to only including adults aged 18 and over, whilst others chose not to place such a restriction on eligible studies. We decided to not place this restriction across the literature identified in this review, due to different age restrictions for purchasing tobacco that exist globally and that some interventions may also target smokers that use tobacco before the local legal age of purchase. There is also small variation in the study designs included in Cochrane reviews; all reviews contain randomised trials (randomised controlled trials), whilst there is some scope to the range of randomised trial designs that are included (such as cluster-randomised, quasi-randomised or crossover trials). Some authors of Cochrane reviews choose to include non-randomised trials in addition to randomised trials, but this often depends on the scope of the research question and the breadth of the literature base (73). We focused only on randomised controlled trials in our study.
Furthermore, hand searching of additional literature was not performed. This may have led to omission of potentially relevant data that could have allowed deeper exploration of the research question. Similarly, only papers accessible via the University of Cambridge medical library and in the English language were used – although it was unlikely that the papers that were inaccessible would have changed our findings, as there was no mention of analysis by SES in study abstracts. A further limitation is that we did not consider the interventions systematically in terms of an available theoretical framework. Explicit use of theory may have allowed further understanding of the kinds of intervention most likely to impact on inequalities in smoking, for example considering if the level of agency required by the patient in interventions delivered by practitioners impacts upon inequalities.
Implications
Currently, trials of smoking cessation interventions delivered in primary care are not designed to allow analysis of effects by measures of SES. These trials generally do not report SES, and there is inconsistency in SES measures in those that do. The majority of studies that do collect data on SES do not analyse by it, and if they did then this analysis would likely be underpowered to assess interaction by SES. Pooling and meta-analysis of data by SES could mitigate this, although this is not currently possible due to infrequent and inconsistent reporting of SES. Our work highlights the need for routine reporting of SES in trials, especially in key areas of inequality such as lifestyle change and multimorbidity, and a greater consensus on measures. Consistent reporting of a core set of SES indicators will enable testing of similarities between trial groups and differential effects by SES. This could include absolute as well as relative measures of SES. Cochrane reviews should consider analysing by SES within their reviews. Responses to these implications should be informed by existing theory and evidence to demonstrate whether an improvement in total health and a reduction in inequality are achieved concurrently.